| Summary: | Large-scale heterogeneous graphs, containing different types of nodes and edges, bring
new challenges to the efficiency and scalability of graph representation learning and
Graph Neural Network-based algorithms. For some graph learning tasks, such as inductive
graph reasoning, pre-training on whole graph data suffers from serious computational
costs. Therefore, This project aims to propose an efficient subgraph extraction
algorithm for large-scale heterogeneous graphs. The extracted subgraph contains a
limited number of representative nodes and selected edge types, which can reflect the
topological characteristics of the entire graph and is expected to be applied to various
graph neural network models.
|
|---|